P-FARFAR2: A multithreaded greedy approach to sampling low-energy RNA structures in Rosetta FARFAR2

Area of Science:

• Computational Biology
• Structural Biology
• Bioinformatics

Background:

• RNA structure prediction is crucial for health science and drug discovery.
• Predicting RNA 3D structures, especially for large sequences, remains a significant challenge.
• Existing methods like Rosetta FARFAR2 show promise but are non-deterministic.

Purpose of the Study:

• To develop P-FARFAR2, a parallel enhancement of the FARFAR2 algorithm.
• To improve the efficiency and accuracy of RNA 3D structure prediction.
• To leverage multithreaded computation for exploring diverse structural configurations.

Main Methods:

• Implemented a parallel mechanism using multithreaded exploration of random configurations.
• Coarsened the synchronization window to multiple Monte Carlo cycles.
• Differentiated threads into one primary and multiple auxiliary threads performing weakened versions of the problem.

Main Results:

• Achieved statistically significant reductions in the energy levels of predicted RNA structures.
• Demonstrated that lower energy levels correlate with improved prediction accuracy.
• Empirically validated the approach on a diverse set of RNA structures.

Conclusions:

• P-FARFAR2 enhances RNA structure prediction by efficiently exploring configurations.
• The parallel approach leads to more accurate and lower-energy RNA models.
• This method offers a significant advancement in computational RNA structure prediction.